Hydropower station abnormal monitoring method and system based on large model data analysis

Abstract

The application provides a hydropower station abnormal monitoring method and system based on large model data analysis, and relates to the technical field of hydropower station operation monitoring. First, multi-dimensional operation information of the whole scene of the hydropower station is collected, including equipment working condition, water flow dynamics, power transmission and environmental interaction information. Then, a pre-trained hydropower station operation coupling large model is used for cross-scene coupling correlation modeling to generate an operation state coupling correlation model. Next, the model and normal operation benchmark data are used to reconstruct an abnormal propagation trajectory. Then, a large model is used to trace the abnormal cause chain. Finally, a dynamically adapted abnormal monitoring response scheme is generated according to the abnormal propagation trajectory and the cause chain and is pushed to a monitoring execution system. The application improves the accuracy and timeliness of hydropower station abnormal monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hydropower station operation monitoring, in particular to a hydropower station abnormal monitoring method and system based on large model data analysis. BACKGROUND

[0002] Traditional hydropower station monitoring methods mainly rely on monitoring and analyzing single equipment or single scene. For example, for the power generation equipment of a hydropower station, only its basic operating parameters such as speed and temperature are usually monitored, and an alarm is issued when these parameters exceed the normal range; for water flow dynamics, indicators such as water level and flow rate are mainly concerned, and a judgment is made according to preset thresholds.

[0003] However, a hydropower station is a complex system engineering, its operation involves equipment working conditions, water flow dynamics, power transmission and environmental interaction, and there is a close coupling relationship between these aspects. The traditional method lacks comprehensive analysis of multi-dimensional operating information and cross-scene coupling correlation modeling capability, and it is difficult to fully and accurately grasp the operating state of the hydropower station. When an abnormal situation occurs, only the surface abnormal phenomenon can be found, the cause chain of the abnormality cannot be traced back in depth, and a targeted monitoring response scheme cannot be timely and effectively formulated, which may lead to further deterioration of the abnormal situation, affect the normal operation of the hydropower station, and even cause safety accidents. SUMMARY

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present application, the present application provides a hydropower station abnormal monitoring method based on large model data analysis, which comprises:

[0005] Collecting multi-dimensional operating information generated by the full-scene operation of the hydropower station, the multi-dimensional operating information including equipment working condition information, water flow dynamic information, power transmission information and environmental interaction information;

[0006] Performing cross-scene coupling correlation modeling on the multi-dimensional operating information by a pre-trained hydropower station operating coupling large model to generate an operating state coupling correlation model;

[0007] Based on the operating state coupling correlation model, combining with normal operating benchmark data of the hydropower station, reconstructing an abnormal propagation trajectory in the operating process of the hydropower station;

[0008] Using the hydropower station operating coupling large model to trace the cause chain of the coupling correlation nodes involved in the abnormal propagation trajectory to generate an abnormal cause chain;

[0009] According to the abnormal propagation trajectory and the abnormal cause chain, generating a dynamically adapted hydropower station abnormal monitoring response scheme, and pushing the hydropower station abnormal monitoring response scheme to a hydropower station monitoring execution system.

[0010] In still another aspect, the present application also provides a hydropower station abnormal monitoring system based on large model data analysis, comprising:

[0011] a processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the above-mentioned hydropower station abnormal monitoring method based on large model data analysis via execution of the machine-executable instructions.

[0012] In still another aspect, the present application also provides a computer program product, which comprises machine-executable instructions stored in a computer-readable storage medium, and a processor of a hydropower station abnormal monitoring system based on large model data analysis reads the machine-executable instructions from the computer-readable storage medium, and the processor executes the machine-executable instructions, so that the hydropower station abnormal monitoring system based on large model data analysis executes the above-mentioned hydropower station abnormal monitoring method based on large model data analysis.

[0013] Based on the above aspects, by collecting multi-dimensional running information generated by the whole scene running of the hydropower station, and using a pre-trained hydropower station running coupling large model to perform cross-scene coupling correlation modeling, a running state coupling correlation model is generated, which can effectively understand the internal relationship and mutual influence between various running elements of the hydropower station. Then, based on the running state coupling correlation model, the abnormal propagation trajectory in the running process of the hydropower station is reconstructed, which can show the path and process of the abnormality from generation to diffusion, so that the operation and maintenance personnel can intuitively understand the influence range and development trend of the abnormality. Using the hydropower station running coupling large model to trace the cause chain of the coupling correlation nodes involved in the abnormal propagation trajectory, an abnormal cause chain is generated, which can deeply mine the root cause of the abnormality. Therefore, a dynamically adapted hydropower station abnormal monitoring response scheme is generated according to the abnormal propagation trajectory and the abnormal cause chain, and the hydropower station abnormal monitoring response scheme is pushed to a hydropower station monitoring execution system, which can realize timely and effective handling of the abnormal situation of the hydropower station, improve the stability and safety of the hydropower station operation, reduce the risk of safety accidents, and ensure the efficient operation of the hydropower station. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 is an execution flow diagram of the hydropower station abnormal monitoring method based on large model data analysis provided by the embodiments of the present application.

[0015] Figure 2 is a schematic diagram of exemplary hardware and software components of the hydropower station abnormal monitoring system based on large model data analysis provided by the embodiments of the present application. DETAILED DESCRIPTION

[0016] The present application will be specifically described below in conjunction with the drawings of the specification, Figure 1It is a flowchart of a water power station abnormal monitoring method based on large model data analysis provided by an embodiment of the present application. The water power station abnormal monitoring method based on large model data analysis will be described in detail below.

[0017] Step S110: Collect multi-dimensional operation information generated by the whole scene operation of the water power station. The multi-dimensional operation information includes equipment working condition information, water flow dynamic information, power transmission information, and environmental interaction information.

[0018] In this embodiment, the collection object is a large basin water power station. The water power station includes multiple water turbine generator units, upstream and downstream reservoirs, multiple power transmission lines, and supporting environmental monitoring equipment. The collection of equipment working condition information covers parameters such as stator winding temperature, rotor winding temperature, bearing temperature, unit vibration acceleration, unit swing, guide vane opening, paddle opening, lubricating oil pressure, lubricating oil temperature, cooling water flow, etc. of the water turbine generator unit; the collection of water flow dynamic information covers parameters such as reservoir water level, inflow, outflow, water flow velocity, water flow pressure, water flow sediment content, tail water level, water head loss, etc.; the collection of power transmission information covers parameters such as phase voltage, line voltage, phase current, line current, power factor, transmission power, line temperature, insulator leakage current, circuit breaker state, etc. of the power transmission line; and the collection of environmental interaction information covers parameters such as air temperature, humidity, wind speed, wind direction, rainfall, snowfall, lightning frequency, soil moisture, water temperature, etc. of the area where the water power station is located.

[0019] During the collection process, for privacy-sensitive data related to equipment operation parameters, data desensitization technology is used to anonymize the identification information of the equipment, such as specific number and geographic location coordinates, replace the equipment number with meaningless character combinations, and convert the geographic location coordinates into relative position information. At the same time, an encrypted transmission method is used to encrypt the data using an asymmetric encryption algorithm during the transmission of the data from the collection terminal to the data processing center, ensuring that the data is not illegally obtained during transmission. When storing, an encrypted storage technology is used to encrypt the data stored in the database, and multiple access permissions are set, so that only authorized technical personnel can access the relevant data after identity verification.

[0020] Step S120: Through a pre-trained water power station operation coupled large model, the multi-dimensional operation information is cross-scene coupled and associated modeled to generate an operation state coupled and associated model.

[0021] Step S1201: The multi-dimensional operation information is split by scene type to form an equipment working condition information set, a water flow dynamic information set, a power transmission information set, and an environmental interaction information set. Each information set contains multiple specific operation records under the same scene.

[0022] In this embodiment, the device working condition information set includes stator winding temperature records of the hydro-generator set A, rotor winding temperature records of the hydro-generator set A, bearing temperature records of the hydro-generator set A, stator winding temperature records of the hydro-generator set B, rotor winding temperature records of the hydro-generator set B, bearing temperature records of the hydro-generator set B, guide vane opening records of the speed governor, paddle opening records of the speed governor, and the like; the water flow dynamic information set includes water level records of the upstream reservoir, water level records of the downstream reservoir, inflow records, outflow records, water flow speed records, water flow pressure records, and the like; the power transmission information set includes phase voltage records of the transmission line C, line voltage records of the transmission line C, phase current records of the transmission line C, line current records of the transmission line C, phase voltage records of the transmission line D, line voltage records of the transmission line D, phase current records of the transmission line D, line current records of the transmission line D, and the like; the environmental interaction information set includes air temperature records, humidity records, wind speed records, wind direction records, rainfall records, snowfall records, and the like.

[0023] Step S1202: Extract the scene features of the operation records in each information set. The scene features of the device working condition information set include device operation parameter change features and device cooperation correlation features. The scene features of the water flow dynamic information set include water flow form change features and water flow-device interaction features. The scene features of the power transmission information set include transmission efficiency change features and transmission link correlation features. The scene features of the environmental interaction information set include environmental factor change features and environmental-device interaction features.

[0024] In this embodiment, the device operating parameter change feature of the device working condition information set is the change of each device operating parameter at different time points, such as the rising or falling trend of the stator winding temperature at consecutive time points, the fluctuation of the vibration acceleration, etc.; the device cooperation and correlation feature is the correlation of the operating parameters between different devices, such as the correlation of the guide vane opening change and the unit speed change, the correlation of the blade opening change and the unit output change, etc. The water flow pattern change feature of the water flow dynamic information set is the change of the water flow velocity and the water flow pressure at different positions, such as the difference between the water flow velocity at the inlet and the water flow velocity at the outlet, the distribution of the water flow pressure in the pipeline, etc.; the water flow and device interaction feature is the effect of the water flow on the hydroelectric generating set, such as the influence of the water head change on the unit output, the influence of the water flow sediment content on the water turbine blade wear, etc. The transmission efficiency change feature of the power transmission information set is the change of the ratio of the transmission power to the input power of the power transmission line, the change of the transmission loss, etc.; the transmission link correlation feature is the power distribution between different power transmission lines, the influence of the line fault on other lines, etc. The environmental factor change feature of the environmental interaction information set is the change of the environmental parameters such as air temperature, humidity, wind speed, etc., such as the change trend of the air temperature within a day, the mutation of the wind speed, etc.; the environmental and device interaction feature is the influence of the environmental parameters on the device operation, such as the influence of the air temperature change on the stator winding temperature, the influence of the wind speed change on the transmission line galloping, etc.

[0025] Step S1203: identifying the coupling correlation dimensions between different information sets, the coupling correlation dimension between the device working condition information set and the water flow dynamic information set is the action response dimension, the coupling correlation dimension between the device working condition information set and the power transmission information set is the load adaptation dimension, the coupling correlation dimension between the water flow dynamic information set and the environmental interaction information set is the influence feedback dimension, and the coupling correlation dimension between the power transmission information set and the environmental interaction information set is the adaptation adjustment dimension.

[0026] In this embodiment, the action response dimension is the influence of the device working condition change on the water flow dynamics and the reaction of the water flow dynamics change on the device working condition, such as the guide vane opening adjustment leading to the change of the water flow velocity, and the water flow velocity change affecting the unit speed; the load adaptation dimension is the influence of the device working condition change on the power transmission load and the adaptation of the device working condition to the power transmission load change, such as the unit output change leading to the change of the power transmission line load, and the power transmission line load change requiring the unit to adjust the output; the influence feedback dimension is the influence of the environmental factor change on the water flow dynamics and the feedback of the water flow dynamics change on the environment, such as the rainfall increase leading to the increase of the reservoir inflow, and the reservoir inflow increase leading to the rise of the reservoir water level, thereby affecting the surrounding environmental humidity; the adaptation adjustment dimension is the influence of the environmental factor change on the power transmission and the adaptation of the power transmission to the environmental change, such as the wind speed increase leading to the aggravation of the transmission line galloping, and the transmission line galloping aggravation requiring the transmission power to be adjusted to avoid failure.

[0027] Step S1204: For each coupling correlation dimension, scene features in the corresponding two information sets are extracted, interaction relationships between the scene features are analyzed, feature indicators capable of representing interaction strength or adaptation degree are extracted, and a feature indicator set corresponding to each dimension is formed.

[0028] Step S12041: For the action-response dimension, device operating parameter change features in the device working condition information set and water flow pattern change features in the water flow dynamic information set are extracted, interaction relationships between the device operating parameter change features and the water flow pattern change features are analyzed, the influence mode of device operation on water flow pattern and the counteraction mode of water flow pattern on device operation are identified, based on the influence mode and the counteraction mode, feature indicators capable of representing the interaction strength of the two are extracted, and an action-response feature indicator set is formed.

[0029] In this embodiment, the device operating parameter change feature is a guide vane opening change feature, and the water flow pattern change feature is a water flow velocity change feature. The interaction relationship between the two is analyzed, the guide vane opening is increased to cause the water flow passage to widen, and the water flow velocity is increased; the water flow velocity is increased to cause the rotational speed of the hydroelectric generating set to increase, and then the guide vane opening needs to be adjusted. Based on the above interaction relationship, the extracted feature indicators include the ratio of the guide vane opening change amount to the water flow velocity change amount, the ratio of the water flow velocity change amount to the unit rotational speed change amount, the difference between the guide vane opening adjustment response time and the water flow velocity change response time, and the like.

[0030] Step S12042: For the load adaptation dimension, device cooperation correlation features in the device working condition information set and transmission efficiency change features in the power transmission information set are extracted, adaptation relationships between the device cooperation correlation features and the transmission efficiency change features are analyzed, the influence law of the device cooperation mode on the transmission efficiency and the adjustment demand of the transmission efficiency change on the device cooperation are identified, based on the influence law and the adjustment demand, feature indicators capable of representing the adaptation degree of the two are extracted, and a load adaptation feature indicator set is formed.

[0031] In this embodiment, the device cooperation correlation feature is a cooperation correlation feature of unit output and guide vane opening, and the transmission efficiency change feature is a transmission efficiency change feature of the power transmission line. The adaptation relationship between the two is analyzed, when the unit output is increased and the guide vane opening is reasonable, the transmission efficiency of the power transmission line is improved; when the transmission efficiency is reduced, the unit is required to adjust the output or the guide vane opening to adapt to the transmission demand. Based on the above adaptation relationship, the extracted feature indicators include the ratio of the unit output change amount to the transmission efficiency change amount, the ratio of the guide vane opening adjustment amount to the transmission efficiency adjustment amount, the difference between the unit output response time and the transmission efficiency response time, and the like.

[0032] Step S12043: For the influence feedback dimension, the water flow and device interaction feature in the water flow dynamic information set and the environment and device interaction feature in the environment interaction information set are extracted, the feedback relationship between the water flow and device interaction feature and the environment and device interaction feature is analyzed, the influence of the environmental factor on the water flow state and the feedback effect of the water flow state change on the environmental action are identified, based on the influence and the feedback effect, a feature index capable of representing the feedback strength of the two is extracted, and an influence feedback feature index set is formed.

[0033] In this embodiment, the water flow and device interaction feature is the action feature of water flow sediment concentration on water turbine blade wear, and the environment and device interaction feature is the action feature of rainfall on reservoir water level. The feedback relationship between the two is analyzed, the increase of rainfall leads to the increase of inflow, the increase of water flow sediment concentration, and the aggravation of water turbine blade wear; the water turbine blade wear leads to the decrease of unit output, and further affects the power generation scheduling of the reservoir, which may lead to the adjustment of the reservoir water level and the feedback influence on the surrounding environment humidity. Based on the above feedback relationship, the extracted feature indexes include the ratio of the rainfall change amount to the water flow sediment concentration change amount, the ratio of the water flow sediment concentration change amount to the blade wear amount, the ratio of the blade wear amount to the unit output change amount, etc.

[0034] Step S12044: For the adaptation adjustment dimension, the transmission link associated feature in the power transmission information set and the environmental factor change feature in the environment interaction information set are extracted, the adjustment relationship between the transmission link associated feature and the environmental factor change feature is analyzed, the influence of environmental change on the transmission link and the adaptation strategy of the transmission link to environmental change are identified, based on the influence and the adaptation strategy, a feature index capable of representing the adjustment adaptation degree of the two is extracted, and an adaptation adjustment feature index set is formed.

[0035] In this embodiment, the transmission link associated feature is the power distribution associated feature of transmission line A and transmission line B, and the environmental factor change feature is the wind speed change feature. The adjustment relationship between the two is analyzed, the increase of wind speed leads to the aggravation of the dancing of transmission line A, which requires adjusting the power distribution of transmission line A and transmission line B, reducing the transmission power of transmission line A and increasing the transmission power of transmission line B; after the power distribution adjustment, the dancing of transmission line A is relieved, and the wind speed change is adapted. Based on the above adjustment relationship, the extracted feature indexes include the ratio of the wind speed change amount to the transmission power change amount of transmission line A, the ratio of the transmission power change amount of transmission line A to the transmission power change amount of transmission line B, and the difference between the wind speed response time and the power distribution adjustment response time.

[0036] Step S12045: Perform feature integration on the action response characteristic index set, retain the core characteristic index, perform feature screening on the load adaptation characteristic index set, retain the index that can reflect the core adaptation relationship, perform feature optimization on the influence feedback characteristic index set, strengthen the representation of the key feedback relationship, perform feature integration on the adaptation adjustment characteristic index set, retain the index that can reflect the core adjustment adaptation relationship, and perform unified format processing on the processed action response characteristic index set, load adaptation characteristic index set, influence feedback characteristic index set and adaptation adjustment characteristic index set, and output the format-unified characteristic index set of each dimension.

[0037] In this embodiment, when the action response characteristic index set is integrated, the variance of each characteristic index is calculated, and the characteristic index with larger variance is retained. These characteristic indexes can more obviously reflect the interaction strength between the device operation and the water flow dynamics. When the load adaptation characteristic index set is screened, the correlation coefficient of each characteristic index and the transmission efficiency is calculated, and the characteristic index with higher correlation coefficient is retained. These characteristic indexes can more accurately reflect the adaptation relationship between the device cooperation and the transmission efficiency. When the influence feedback characteristic index set is optimized, each characteristic index is weighted processed, the weight of the characteristic index related to the blade wear is increased, and the representation of the key feedback relationship is strengthened. When the adaptation adjustment characteristic index set is integrated, the information gain of each characteristic index is calculated, and the characteristic index with larger information gain is retained. These characteristic indexes can more effectively reflect the adjustment adaptation relationship between the transmission link and the environmental change. When the unified format processing is performed, the numerical range of each characteristic index is converted to the same interval, and the name of the characteristic index is converted to the unified naming rule, such as adopting the naming method of “feature type-feature serial number”.

[0038] Step S1205: Perform redundant information elimination processing on each characteristic index set, retain the core characteristic index that can reflect the core interaction relationship, and input all core characteristic indexes into the factor generation module of the hydropower station operation coupling large model to generate the coupling correlation factor corresponding to each coupling correlation dimension. The coupling correlation factor is directly related to the core characteristic index of the corresponding dimension.

[0039] In this embodiment, the correlation analysis method is adopted for the redundant information elimination processing, the correlation coefficient between any two characteristic indexes in the characteristic index set is calculated, when the correlation coefficient is greater than the preset threshold, it is considered that these two characteristic indexes have redundant information, and one of them is retained. For example, in the action response characteristic index set, the correlation coefficient between the ratio of guide vane opening change amount and water flow velocity change amount and the ratio of water flow velocity change amount and unit speed change amount is greater than the preset threshold, and the ratio of guide vane opening change amount and water flow velocity change amount is retained.

[0040] The all core feature indicators are input into a factor generation module of a hydropower station operation coupling large model, the factor generation module includes multiple fully connected layers, the core feature indicators are first input into a first fully connected layer, after being processed by an activation function, an intermediate feature vector is output; the intermediate feature vector is input into a second fully connected layer, after being processed by an activation function, a coupling correlation factor is output. The dimension of the coupling correlation factor is related to the dimension of the core feature indicators, each coupling correlation factor corresponds to a core feature indicator, and reflects the action strength of the core feature indicator in the coupling correlation dimension.

[0041] Step S1206: input all coupling correlation factors into an association modeling module of the hydropower station operation coupling large model, the association modeling module is based on the deep semantic understanding ability of the large model, and constructs an association logic network between multiple-dimension operation information, in the association logic network, each operation record is taken as a network node, each coupling correlation factor is taken as a connecting link between nodes, and an initial coupling correlation framework is formed.

[0042] In the embodiment, the association modeling module includes a graph neural network layer, the coupling correlation factors are first input into the graph neural network layer, the graph neural network layer processes the coupling correlation factors, and the connection relationship between nodes is constructed. Each operation record in the device working condition information set is taken as a node, such as a stator winding temperature record node of a hydro-generator unit A, a rotor winding temperature record node of the hydro-generator unit A and the like; each operation record in the water flow dynamic information set is taken as a node, such as a reservoir water level record node, an inflow record node and the like; each operation record in the power transmission information set is taken as a node, such as a phase voltage record node of a power transmission line A, a line current record node of the power transmission line A and the like; and each operation record in the environmental interaction information set is taken as a node, such as a temperature record node, a humidity record node and the like.

[0043] Each coupling correlation factor is taken as a connecting link between nodes, such as a coupling correlation factor in the action response dimension connecting a device working condition information node and a water flow dynamic information node, a coupling correlation factor in the load adaptation dimension connecting a device working condition information node and a power transmission information node, a coupling correlation factor in the influence feedback dimension connecting a water flow dynamic information node and an environmental interaction information node, and a coupling correlation factor in the adaptive adjustment dimension connecting a power transmission information node and an environmental interaction information node. The initial coupling correlation framework formed includes multiple nodes and the connection relationship between the nodes, and each connection relationship corresponds to a coupling correlation factor.

[0044] Step S1207: Iterative optimization is performed on the initial coupling correlation framework by a dynamic optimization module of the hydropower station operation coupling large model, the correlation strength between nodes is adjusted, and the optimized correlation logic network and the scene characteristics of each information set are integrated to generate an operation state coupling correlation model capable of reflecting the cross-scene interaction relationship of multi-dimensional operation information.

[0045] In this embodiment, the dynamic optimization module includes an iterative optimization algorithm. First, the error of the correlation strength between each node in the initial coupling correlation framework is calculated. The error is obtained by comparing the initial correlation strength with the correlation strength in the actual operation data. Then, the correlation strength between nodes is adjusted according to the error, the correlation strength with smaller error is increased, and the correlation strength with larger error is reduced. At the same time, the correlation logic between nodes is adjusted. When the correlation logic between nodes in the actual operation data does not match the initial correlation logic, the initial correlation logic is modified.

[0046] The iterative optimization process is repeated until the error is less than a preset threshold. For example, in the initial coupling correlation framework, the correlation strength error between the stator winding temperature record node of the hydro-generator unit A and the reservoir water level record node is large. By adjusting the correlation strength, it is consistent with the correlation strength in the actual operation data. When adjusting the correlation logic, when the actual operation data shows that the increase of the stator winding temperature of the hydro-generator unit A will cause the guide vane opening to decrease, and the initial correlation logic shows that the increase of the stator winding temperature of the hydro-generator unit A will cause the guide vane opening to increase, the initial correlation logic is modified to the increase of the stator winding temperature of the hydro-generator unit A will cause the guide vane opening to decrease.

[0047] When integrating the optimized correlation logic network and the scene characteristics of each information set, the device operation parameter change characteristics and the water flow form change characteristics in the scene characteristics are associated with the nodes in the correlation logic network, the characteristic indexes in the scene characteristics are taken as the attributes of the nodes, and the operation state coupling correlation model is generated. The operation state coupling correlation model includes nodes, node attributes, correlation strength between nodes, and correlation logic, and can reflect the cross-scene interaction relationship of multi-dimensional operation information.

[0048] Step S1208: The operation state coupling correlation model is adjusted for scene adaptation, and the adjusted operation state coupling correlation model is output.

[0049] In this embodiment, the scene adaptation adjustment is based on the actual operation scene of the hydropower station, considering actual parameters such as the number of units of the hydropower station, the reservoir capacity, the length of the power transmission line, etc. For example, when the number of units of the hydropower station increases, the corresponding device working condition information node is added in the operation state coupling correlation model, and the correlation strength and correlation logic between nodes are adjusted to adapt to the operation scene after the number of units increases. When the reservoir capacity decreases, the correlation strength between the water flow dynamic information node and the device working condition information node is adjusted to adapt to the water flow change after the reservoir capacity decreases. After the scene adaptation adjustment is completed, the adjusted operation state coupling correlation model is output.

[0050] Step S130: Based on the operation state coupling correlation model, the abnormal propagation trajectory in the operation process of the hydropower station is reconstructed in combination with the normal operation benchmark data of the hydropower station.

[0051] Step S1301: Collect multi-dimensional historical operation information accumulated in the long-term normal operation stage of the hydropower station.

[0052] In this embodiment, the long-term normal operation stage is a time period during which the hydropower station is continuously operated and no abnormal event occurs, and the multi-dimensional historical operation information includes device working condition information, water flow dynamic information, power transmission information and environmental interaction information in this time period. The time span of the collected historical operation information is several years, ensuring that the operation data under different seasons and different weather conditions are included.

[0053] Step S1302: The multi-dimensional historical operation information is split by scene type to form a historical device working condition information set, a historical water flow dynamic information set, a historical power transmission information set and a historical environmental interaction information set.

[0054] In this embodiment, the historical device working condition information set includes the stator winding temperature record of the hydro-generator unit A in the normal operation stage, the rotor winding temperature record of the hydro-generator unit A in the normal operation stage, etc.; the historical water flow dynamic information set includes the water level record of the reservoir in the normal operation stage, the inflow record, etc.; the historical power transmission information set includes the phase voltage record of the power transmission line C in the normal operation stage, the line current record of the power transmission line C in the normal operation stage, etc.; and the historical environmental interaction information set includes the air temperature record of the area where the hydropower station is located in the normal operation stage, the humidity record, etc.

[0055] Step S1303: Extract the historical scene features of the running records in each historical information set. The historical scene features of the historical equipment working condition information set include historical equipment running parameter change features and historical equipment cooperation correlation features. The historical scene features of the historical water flow dynamic information set include historical water flow form change features and historical water flow and equipment interaction features. The historical scene features of the historical power transmission information set include historical transmission efficiency change features and historical transmission link correlation features. The historical scene features of the historical environment interaction information set include historical environment factor change features and historical environment and equipment interaction features.

[0056] In this embodiment, the historical equipment running parameter change features are the change trend of the stator winding temperature of the hydroelectric generating set A in the normal running stage. For example, when the air temperature is relatively high in summer, the stator winding temperature shows a slow upward trend. The historical equipment cooperation correlation features are the cooperation relationship between the guide vane opening and the paddle opening of the hydroelectric generating set A in the normal running stage. For example, when the guide vane opening increases, the paddle opening also increases accordingly. The historical water flow form change features are the change trend of the reservoir water level in the normal running stage. For example, when the inflow increases in the rainy season, the reservoir water level shows an upward trend. The historical water flow and equipment interaction features are the influence of the water flow sediment content on the wear of the water turbine blade in the normal running stage. For example, when the water flow sediment content is low, the wear degree of the blade is small. The historical transmission efficiency change features are the transmission efficiency change trend of the power transmission line C in the normal running stage. For example, in the power consumption peak period, the transmission efficiency shows a downward trend. The historical transmission link correlation features are the power distribution relationship between the power transmission line C and the power transmission line D in the normal running stage. For example, when the transmission power of the power transmission line C is large, the transmission power of the power transmission line D is small. The historical environment factor change features are the change trend of the air temperature in the normal running stage. For example, within a day, the air temperature shows an upward trend from morning to noon. The historical environment and equipment interaction features are the influence of the wind speed on the dancing of the power transmission line in the normal running stage. For example, when the wind speed is small, the dancing degree of the power transmission line is small.

[0057] Step S1304: Identify the historical coupling correlation dimensions between different historical information sets. The historical coupling correlation dimension between the historical equipment working condition information set and the historical water flow dynamic information set is the historical action response dimension. The historical coupling correlation dimension between the historical equipment working condition information set and the historical power transmission information set is the historical load adaptation dimension. The historical coupling correlation dimension between the historical water flow dynamic information set and the historical environment interaction information set is the historical influence feedback dimension. The historical coupling correlation dimension between the historical power transmission information set and the historical environment interaction information set is the historical adaptation adjustment dimension.

[0058] In this embodiment, the historical action response dimension is the influence of historical equipment working condition changes on historical water flow dynamics and the counteraction of historical water flow dynamics changes on historical equipment working conditions, such as historical guide vane opening adjustment leading to historical water flow speed changes, and historical water flow speed changes affecting historical unit rotation speed; the historical load adaptation dimension is the influence of historical equipment working condition changes on historical power transmission load and the adaptation of historical power transmission load changes to historical equipment working conditions, such as historical unit output changes leading to historical power transmission line load changes, and historical power transmission line load changes requiring historical unit output adjustment; the historical influence feedback dimension is the influence of historical environmental factor changes on historical water flow dynamics and the feedback of historical water flow dynamics changes to historical environmental action, such as historical rainfall increase leading to historical reservoir inflow increase, and historical reservoir inflow increase leading to historical reservoir water level rise, which feeds back to affect the surrounding environmental humidity; the historical adaptation adjustment dimension is the influence of historical environmental factor changes on historical power transmission and the adaptation of historical power transmission to historical environmental changes, such as historical wind speed increase leading to historical power transmission line dancing intensification, and historical power transmission line dancing intensification requiring historical power transmission power adjustment to avoid failure.

[0059] Step S1305: For each historical coupling correlation dimension, a corresponding historical coupling correlation factor is generated, which is determined based on the interaction relationship of historical scene features in different historical information sets.

[0060] In this embodiment, the generation method of the historical coupling correlation factor is the same as that of the coupling correlation factor in step S125. First, the historical scene features are extracted and processed, then the redundant information elimination processing is performed, the core historical scene features are retained, and finally the core historical scene features are input into the factor generation module of the hydropower station operation coupling large model to generate the historical coupling correlation factor. For example, in the historical action response dimension, the core historical scene feature is the ratio of the historical guide vane opening change amount to the historical water flow speed change amount. The core historical scene feature is input into the factor generation module to generate the coupling correlation factor of the historical action response dimension.

[0061] Step S1306: All historical coupling correlation factors are input into the correlation modeling module of the hydropower station operation coupling large model to build a historical correlation logic network between multi-dimensional historical operation information.

[0062] In this embodiment, the processing procedure of the association modeling module is the same as that in step S126, the historical coupling association factors are input into the graph neural network layer to construct a historical association logic network. The nodes in the historical association logic network are historical operation records, and the connection links between the nodes are historical coupling association factors. For example, the stator winding temperature record node of the hydro-generator unit A in the historical equipment working condition information set is connected with the reservoir water level record node in the historical water flow dynamic information set through the coupling association factor of the historical action response dimension.

[0063] Step S1307: In the historical association logic network, each historical operation record is taken as a historical network node, and each historical coupling association factor is taken as a connection link between historical nodes to form a historical initial coupling association framework.

[0064] In this embodiment, the formation process of the historical initial coupling association framework is the same as that in step S126, the historical operation records are taken as historical network nodes, and the historical coupling association factors are taken as connection links between historical nodes to construct a historical initial coupling association framework. The historical initial coupling association framework includes historical nodes, connection relationships between historical nodes, and historical coupling association factors.

[0065] Step S1308: The dynamic optimization module of the hydropower station operation coupling large model is used to iteratively optimize the historical initial coupling association framework, adjust the association strength and association logic between historical nodes, and eliminate historical association conflicts and historical redundant associations.

[0066] In this embodiment, the processing procedure of the dynamic optimization module is the same as that in step S127, the error of the association strength between each historical node in the historical initial coupling association framework is first calculated, then the association strength between historical nodes is adjusted according to the error, the association logic between historical nodes is adjusted, and historical association conflicts and historical redundant associations are eliminated. The iterative optimization process is repeated until the error is less than a preset threshold.

[0067] Step S1309: The optimized historical association logic network and the historical scene features of each historical information set are integrated to generate a plurality of normal operation state coupling association models, common association logic, stable coupling factors, and node interaction rules in all normal operation state coupling association models are extracted, and a hydropower station normal operation benchmark data is integrated and formed.

[0068] In this embodiment, when integrating the optimized historical correlation logic network and the historical scene features of each historical information set, the historical equipment operation parameter change features and the historical water flow form change features in the historical scene features are associated with the historical nodes in the historical correlation logic network to generate a plurality of normal operation state coupling correlation models. When extracting common correlation logic, the correlation logic in the plurality of normal operation state coupling correlation models is analyzed to find the same correlation logic. When extracting stable coupling factors, the coupling factors in the plurality of normal operation state coupling correlation models are analyzed to find the coupling factors with smaller numerical changes. When extracting node interaction rules, the node interaction in the plurality of normal operation state coupling correlation models is analyzed to find the same interaction rules. When integrating the normal operation benchmark data of the hydropower station, the common correlation logic, the stable coupling factors and the node interaction rules are combined together to form structured benchmark data.

[0069] Step S13010: comprehensively comparing the current generated operation state coupling correlation model with the normal operation benchmark data of the hydropower station to identify deviation correlation nodes and deviation coupling factors in the operation state coupling correlation model that deviate from the normal operation benchmark data.

[0070] In this embodiment, the comprehensive comparison adopts a feature matching method to compare the node attributes, the correlation strength between nodes and the coupling factors in the operation state coupling correlation model with the corresponding contents in the normal operation benchmark data. When the difference in the node attributes is greater than a preset threshold, the node is considered as a deviation correlation node. When the difference in the correlation strength is greater than a preset threshold, the coupling factor corresponding to the correlation strength is considered as a deviation coupling factor. For example, the temperature value of the stator winding temperature record node of the hydro-generator unit A in the operation state coupling correlation model is greatly different from the corresponding temperature value in the normal operation benchmark data, and the node is a deviation correlation node. The coupling factor corresponding to the ratio of the guide vane opening change amount to the water flow velocity change amount in the operation state coupling correlation model is greatly different from the corresponding coupling factor in the normal operation benchmark data, and the coupling factor is a deviation coupling factor.

[0071] Step S13011: tracing the original multidimensional operation information corresponding to each deviation correlation node to extract the operation state features and the correlation interaction features of the deviation correlation node.

[0072] In this embodiment, when tracing the original multidimensional operation information, the corresponding original operation record is found in the multidimensional operation information database through the identification information of the deviation correlation node. When extracting the operation state features, the equipment operation parameters, the water flow parameters, the power transmission parameters and the environmental parameters in the original operation record are extracted. When extracting the correlation interaction features, the interaction record of the node and other nodes in the original operation record is extracted, such as the record of the parameter change of the node leading to the parameter change of other nodes.

[0073] Step S13012: analyze the coupling association relationship between the deviation association nodes, and deduce the connection path and interaction sequence between the deviation association nodes based on the association logic network in the running state coupling association model.

[0074] Step S130121: extract the node attributes and association identifiers of all the deviation association nodes from the association logic network in the running state coupling association model.

[0075] In this embodiment, the node attributes include the device type, parameter type, and value range of the deviation association nodes; and the association identifiers include the connection relationship identifier of the deviation association nodes and other nodes, coupling factor identifier, and the like. For example, the deviation association node is a stator winding temperature record node of a hydro-generator unit A, the node attributes are that the device type is a hydro-generator unit, the parameter type is a stator winding temperature, and the value range is a normal operation range; and the association identifiers are the connection relationship identifier of the stator winding temperature record node and the guide vane opening record node of the hydro-generator unit A, the connection relationship identifier of the stator winding temperature record node and the water flow speed record node, and the corresponding coupling factor identifier.

[0076] Step S130122: determine the position of each deviation association node in the association logic network and the adjacent association nodes based on the node attributes and association identifiers, wherein the adjacent association nodes include the directly associated normal nodes and deviation nodes.

[0077] In this embodiment, when the position of the deviation association node in the association logic network is determined, the coordinate position of the node in the association logic network is found according to the node attributes and association identifiers; and when the adjacent association nodes are determined, the nodes directly connected to the deviation association node are found, including the normal nodes and deviation nodes. For example, the position of the stator winding temperature record node of the hydro-generator unit A in the association logic network is a coordinate (x1, y1), and the adjacent association nodes thereof are the guide vane opening record node of the hydro-generator unit A (normal node), the rotor winding temperature record node of the hydro-generator unit A (deviation node), and the water flow speed record node (deviation node).

[0078] Step S130123: analyze the coupling association type between each deviation association node and the adjacent association nodes, and distinguish the action response type association, load adaptation type association, influence feedback type association, and adaptation adjustment type association.

[0079] In this embodiment, the distinction of the coupling correlation type is based on the interaction relationship between nodes. The action response type correlation is the correlation between the equipment working condition node and the water flow dynamic node, such as the correlation between the stator winding temperature record node and the water flow velocity record node. The load adaptation type correlation is the correlation between the equipment working condition node and the power transmission node, such as the correlation between the stator winding temperature record node and the power transmission line power record node. The influence feedback type correlation is the correlation between the water flow dynamic node and the environmental interaction node, such as the correlation between the water flow velocity record node and the air temperature record node. The adaptive adjustment type correlation is the correlation between the power transmission node and the environmental interaction node, such as the correlation between the power transmission line power record node and the wind speed record node.

[0080] Step S130124: For each deviation correlation node, trace the interaction history between it and the adjacent correlation node, extract the data transmission record, the state change record and the coupling factor change record in the interaction process.

[0081] In this embodiment, the tracing of the interaction history is achieved by searching the interaction record between the deviation correlation node and the adjacent correlation node in the multidimensional running information database. The data transmission record includes the data content, the transmission time, the transmission mode and the like transmitted between nodes. The state change record includes the change time, the change amplitude, the change trend and the like of the node parameter. The coupling factor change record includes the change time, the change amplitude, the change trend and the like of the coupling factor. For example, the interaction history between the stator winding temperature record node of the hydroelectric generating set A and the guide vane opening record node of the hydroelectric generating set A is traced, and the extracted data transmission record is the time of the stator winding temperature data transmission to the guide vane opening control module, the transmission content and the like. The extracted state change record is the time and amplitude of the decrease of the guide vane opening caused by the increase of the stator winding temperature. The extracted coupling factor change record is the change time and amplitude of the coupling factor corresponding to the ratio of the guide vane opening change amount to the water flow velocity change amount.

[0082] Step S130125: Based on the interaction history record, the interaction cause-effect relationship between the deviation correlation node and the adjacent correlation node is deduced, and the trigger condition and the transmission logic of the deviation propagation are determined.

[0083] In this embodiment, the derivation of the interactive causal relationship adopts a causal analysis method to analyze the time sequence and logical relationship among the data transmission records, state change records and coupling factor change records in the interaction history records. When the state change of one node occurs before the state change of another node, and there is a data transmission record between the two, the former is considered to be the cause and the latter is considered to be the effect. The trigger condition is the initial condition that causes the deviation propagation, such as the stator winding temperature rising to a preset threshold; the transmission logic is the way in which the deviation propagates from one node to another, such as the stator winding temperature rising causing the guide vane opening to decrease, the guide vane opening decreasing causing the water flow velocity to decrease, and the water flow velocity decreasing causing the unit rotation speed to decrease.

[0084] Step S130126: Starting from the initial deviation node, the deviation nodes directly associated with the initial deviation node are sequentially derived according to the causal relationship and the coupling association type, forming a first-level propagation path; taking the deviation nodes in the first-level propagation path as the starting point, the subsequent deviation nodes associated therewith are continuously derived, forming a second-level propagation path, and the nodes are sequentially derived in this way until all the associated deviation nodes are covered, and the node sequence, coupling association type and interaction time node in each propagation path are recorded, forming a preliminary path sequence.

[0085] In this embodiment, the initial deviation node is the first node that appears deviation, such as the stator winding temperature record node. Starting from the initial deviation node, the deviation nodes directly associated with the initial deviation node are derived according to the interactive causal relationship and the coupling association type, such as the guide vane opening record node, the rotor winding temperature record node, etc., forming a first-level propagation path. Taking the guide vane opening record node in the first-level propagation path as the starting point, the subsequent deviation nodes associated therewith are derived, such as the water flow velocity record node, the unit rotation speed record node, etc., forming a second-level propagation path. The nodes are sequentially derived in this way until all the associated deviation nodes are covered. The node sequence, coupling association type and interaction time node in each propagation path are recorded, such as the node sequence of the first-level propagation path being the stator winding temperature record node→guide vane opening record node→rotor winding temperature record node, the coupling association type being the action-response type association, and the interaction time node being t1→t2→t3.

[0086] Step S130127: The preliminary path sequence is subjected to a deduplication processing, the interaction sequence of each node in the deduplicated path sequence is analyzed to determine whether it fits the preset interaction rules in the association logic network, the node sequence that does not fit is adjusted, and all the path sequences that fit the rules are integrated, forming a complete connection path and interaction sequence among the deviation association nodes.

[0087] In this embodiment, the deduplication processing adopts a path matching method to delete the same paths in the preliminary path sequence and retain the unique paths. When analyzing whether the interaction sequence matches the preset interaction rule, the node interaction sequence in the path sequence is compared with the preset interaction rule in the associated logical network. When the interaction sequence does not match the preset interaction rule, the node sequence is adjusted. For example, the node sequence in the preliminary path sequence is stator winding temperature recording node→rotor winding temperature recording node→guide vane opening recording node, and the preset interaction rule in the associated logical network is stator winding temperature recording node→guide vane opening recording node→rotor winding temperature recording node. The adjusted node sequence is stator winding temperature recording node→guide vane opening recording node→rotor winding temperature recording node. When integrating the path sequences that match the rules, the adjusted path sequences are combined together to form the complete connection path and the interaction sequence.

[0088] Step S130128: mapping the complete connection path and the interaction sequence to the associated logical network, marking the specific position and the associated relationship of the path in the network, based on the marked mapping result, checking the accuracy of the connection path and the interaction sequence again, correcting the path deviation found in the mapping process, and outputting the derived complete connection path and the interaction sequence.

[0089] In this embodiment, the mapping process corresponds the nodes in the complete connection path and the interaction sequence to the nodes in the associated logical network, marks the specific position of the path in the network, such as the starting node position, the ending node position, and the passing node position of the path in the network, and marks the associated relationship of the path with other paths, such as the intersection point and the connection point of the path with other paths. When checking the accuracy again, it is checked whether the mapped path is consistent with the node connection relationship in the associated logical network, and whether there is a path interruption, a path error, or the like. When correcting the path deviation, when it is found that the path is not consistent with the node connection relationship in the associated logical network, the node sequence or the node connection relationship of the path is adjusted.

[0090] Step S13013: identifying the coupling associated dimension corresponding to the deviation coupling factor, and determining the propagation direction and the influence range of the deviation in different coupling associated dimensions.

[0091] In this embodiment, when identifying the coupling correlation dimension, the corresponding coupling correlation dimension is determined according to the name or identification information of the deviation coupling factor. For example, the deviation coupling factor is the ratio of the guide vane opening change amount to the water flow velocity change amount, and the corresponding coupling correlation dimension is the action response dimension. When determining the propagation direction, the propagation direction of the deviation is determined according to the action direction of the deviation coupling factor in the coupling correlation dimension, such as the deviation coupling factor in the action response dimension causing the deviation to propagate from the equipment working condition node to the water flow dynamic node. When determining the influence range, the influence range of the deviation in the coupling correlation dimension is determined according to the numerical value and the action strength of the deviation coupling factor, such as the influence range being wider when the numerical value of the deviation coupling factor is larger.

[0092] Step S13014: based on the connection path, the interaction sequence and the deviation propagation direction of the deviation correlation node, an initial abnormal propagation path framework is constructed.

[0093] In this embodiment, the construction of the initial abnormal propagation path framework combines the connection path, the interaction sequence and the deviation propagation direction together to form a structured path framework. The framework includes the starting node, the terminal node, the passing node, the interaction sequence between the nodes, the deviation propagation direction and the like. For example, the starting node of the initial abnormal propagation path framework is the stator winding temperature recording node, the terminal node is the unit rotating speed recording node, the passing node is the guide vane opening recording node and the water flow velocity recording node, the interaction sequence is the stator winding temperature recording node→the guide vane opening recording node→the water flow velocity recording node→the unit rotating speed recording node, and the deviation propagation direction is from the equipment working condition node to the water flow dynamic node, and then to the equipment working condition node.

[0094] Step S13015: after supplementing the deviation degree, the interaction time sequence and the coupling factor change of each node in the initial abnormal propagation path framework, the initial abnormal propagation path framework is input into the trajectory reconstruction module of the hydropower station operation coupling large model, the initial abnormal propagation path framework is logically checked and structurally optimized, and the node interaction logic in the path is adjusted to be consistent with the coupling correlation model.

[0095] Step S130151: the initial abnormal propagation path framework is input into the trajectory reconstruction module of the hydropower station operation coupling large model, the correlation logic rules in the operation state coupling correlation model are extracted, including the node interaction type rule, the coupling correlation dimension rule and the propagation sequence rule, each node interaction link in the initial path framework is checked one by one to see whether it conforms to the correlation logic rule, and the abnormal interaction link that does not conform to the rule is identified.

[0096] In this embodiment, the node interaction type rule is that the interaction type between nodes should match the coupling correlation dimension, such as the node interaction type in the action-response dimension should be the interaction between the device working condition node and the water flow dynamic node; the coupling correlation dimension rule is that the coupling factor should match the coupling correlation dimension, such as the coupling factor in the action-response dimension should be the ratio between the device working condition parameter and the water flow dynamic parameter; the propagation order rule is that the order of deviation propagation should comply with the logical order of device operation, such as the increase of stator winding temperature should first cause the decrease of guide vane opening, and then cause the decrease of water flow speed. The initial path framework is checked one by one to see whether each node interaction link in the initial path framework complies with the above rules. When the node interaction type does not match the coupling correlation dimension, the interaction link is an abnormal interaction link; when the coupling factor does not match the coupling correlation dimension, the interaction link is an abnormal interaction link; when the propagation order does not comply with the logical order, the interaction link is an abnormal interaction link.

[0097] Step S130152: For each abnormal interaction link, the node attribute, coupling correlation factor and interaction time information of the abnormal interaction link are extracted, the reason why the abnormal interaction link does not comply with the rules is analyzed, the node interaction mode or interaction order is adjusted in combination with the node correlation relationship in the correlation logic network.

[0098] In this embodiment, when the node attribute, coupling correlation factor and interaction time information of the abnormal interaction link are extracted, the device type, parameter type, value range and other attributes of the node are obtained, the name, value and other information of the coupling correlation factor are obtained, and the start time, end time and other information of the interaction time are obtained. When the reason why the rules are not complied with is analyzed, the rule violated by the abnormal interaction link is found out according to the node interaction type rule, the coupling correlation dimension rule and the propagation order rule. For example, the abnormal interaction link is the interaction between the stator winding temperature recording node and the power transmission line power recording node, which violates the node interaction type rule because the interaction type should be load adaptation type correlation, while the actual interaction type is action-response type correlation. When the node interaction mode or interaction order is adjusted, the interaction type is adjusted from action-response type correlation to load adaptation type correlation, or the node interaction order is adjusted to stator winding temperature recording node→guide vane opening recording node→power transmission line power recording node.

[0099] Step S130153: It is checked whether the application of the coupling correlation factor in the initial path framework matches the coupling correlation dimension, the coupling correlation factor that does not match is adjusted, it is checked whether the propagation order of the node in the initial path framework complies with the actual change time sequence of the multi-dimensional operation information, the node order that is in time sequence disorder is corrected, after the logical verification is completed, the node distribution density of the adjusted path framework is analyzed, the node links with a density greater than a set density threshold are merged, the optimized path framework is mapped and compared with the operation state coupling correlation model, based on the mapping and comparison result, the path framework is finally adjusted, and the abnormal propagation trajectory is output.

[0100] In this embodiment, when checking whether the application of the coupling correlation factor matches the coupling correlation dimension, the name or identification information of the coupling correlation factor is compared with the coupling correlation dimension, and when they do not match, the coupling correlation factor is adjusted to match the coupling correlation dimension. When checking whether the propagation order of the nodes matches the actual change timing, the interaction time of the nodes is compared with the actual time in the multi-dimensional operation information, and when the timing is disordered, the node order is corrected. When analyzing the node distribution density, the number of nodes per unit length in the path framework is calculated, and when the number is greater than the set density threshold, adjacent node links are merged. When mapping and comparing, the optimized path framework is compared with the node connection relationship in the operation state coupling correlation model, and when the node connection relationship in the path framework does not match the node connection relationship in the model, the node connection relationship of the path framework is adjusted.

[0101] Step S13016: According to the optimized path framework, the abnormal propagation trajectory is reconstructed, which includes the initial deviation node, the intermediate propagation node, the terminal impact node and the propagation logic chain between nodes.

[0102] In this embodiment, the initial deviation node is the first node that deviates, such as the stator winding temperature record node; the intermediate propagation node is the node passed through during the propagation of the deviation from the initial deviation node to the terminal impact node, such as the guide vane opening record node and the water flow velocity record node; and the terminal impact node is the final node of the deviation propagation, such as the unit speed record node. The propagation logic chain is the interaction logic and propagation mode between nodes, such as the stator winding temperature rising leading to the guide vane opening decreasing, the guide vane opening decreasing leading to the water flow velocity decreasing, and the water flow velocity decreasing leading to the unit speed decreasing.

[0103] Step S13017: Feature labeling is performed on each propagation link in the abnormal propagation trajectory to determine the coupling correlation dimension, the deviation change trend and the node interaction mode of each link, and the labeled abnormal propagation trajectory is output.

[0104] In this embodiment, when feature labeling is performed, the coupling correlation dimension of each propagation link is determined, such as the coupling correlation dimension of the propagation link from the stator winding temperature record node to the guide vane opening record node being the action response dimension; the deviation change trend is determined, such as the stator winding temperature rising and the guide vane opening decreasing; and the node interaction mode is determined, such as data transmission and parameter adjustment. When the labeled abnormal propagation trajectory is output, the labeling information is added to the abnormal propagation trajectory to form structured trajectory data.

[0105] Step S140: The abnormal cause chain is generated by using the hydropower station operation coupling large model to trace the cause chain of the coupling correlation nodes involved in the abnormal propagation trajectory.

[0106] Step S1401: Extract all coupling correlation nodes in the abnormal propagation track to form an abnormal correlation node set, the coupling correlation nodes including initial deviation nodes, intermediate propagation nodes and terminal impact nodes.

[0107] In this embodiment, the abnormal correlation node set includes a stator winding temperature record node, a guide vane opening record node, a water flow velocity record node, a unit rotation speed record node, etc.

[0108] Step S1402: Collect multi-dimensional running information original records and scene features corresponding to each abnormal correlation node to form a node correlation information set.

[0109] In this embodiment, the node correlation information set includes original temperature data of the stator winding temperature record node, original opening data of the guide vane opening record node, original speed data of the water flow velocity record node, original rotation speed data of the unit rotation speed record node, etc., and corresponding scene features, such as the scene features of the stator winding temperature record node, which are stator winding temperature change trend and temperature distribution, etc.

[0110] Step S1403: Input the abnormal correlation node set and the node correlation information set into a cause tracing module of the hydropower station running coupling large model, call a preset hydropower station abnormal cause knowledge base, the hydropower station abnormal cause knowledge base including abnormal correlation nodes, propagation track segments and complete cause records corresponding to historical abnormal events, the cause records including direct causes, indirect causes and cause interaction relationships.

[0111] In this embodiment, the hydropower station abnormal cause knowledge base includes abnormal correlation nodes, propagation track segments and cause records corresponding to a stator winding temperature too high event occurred in history, such as that the direct cause of the stator winding temperature too high is a cooling system failure, the indirect cause is an environment temperature too high, and the cause interaction relationship is that the environment temperature too high aggravates the influence of the cooling system failure.

[0112] Step S1404: Through a semantic matching module of the hydropower station running coupling large model, deeply match scene features of the current abnormal correlation node with scene features of historical abnormal correlation nodes in the knowledge base, filter out historical abnormal events with scene feature matching degrees meeting requirements, and extract cause records and propagation track segments corresponding thereto.

[0113] In this embodiment, the semantic matching module adopts a semantic similarity calculation method to convert the scene features of the current abnormal association node and the scene features of the historical abnormal association node in the knowledge base into semantic vectors, and calculate the similarity between the semantic vectors. When the similarity is greater than a preset threshold, it is considered that the scene feature matching degree meets the requirements, and the corresponding historical abnormal event is selected. When extracting the cause record and the propagation track segment, the direct cause, the indirect cause, the cause interaction relationship and the propagation track segment of the historical abnormal event are obtained.

[0114] Step S1405: Comparing and analyzing the historical cause record with the node interaction logic and coupling association type of the current abnormal propagation track, identifying the reusable cause association clues, and based on the cause association clues, combining the running information original record of the current abnormal association node, preliminarily inferring the potential cause corresponding to each abnormal association node.

[0115] In this embodiment, when comparing and analyzing the historical cause record with the node interaction logic and coupling association type of the current abnormal propagation track, the same or similar parts between the two are found. The reusable cause association clues are the cause information related to the current abnormal propagation track in the historical cause record, such as the cooling system failure in the historical abnormal event causing the temperature of the stator winding to be too high, and the temperature of the stator winding in the current abnormal propagation track is also too high, the cooling system failure is a reusable cause association clue. When preliminarily inferring the potential cause, the reusable cause association clue is combined with the running information original record of the current abnormal association node, such as the cooling system flow data anomaly of the current abnormal association node, and the potential cause is inferred as cooling system failure.

[0116] Step S1406: Verifying the cause association of each potential cause, and analyzing the fit of the potential cause with the abnormal node state change and the inter-node propagation logic.

[0117] Step S14061: Extracting the cause feature corresponding to each potential cause, which includes the cause object, the action mode, the action timing and the expected impact effect.

[0118] In this embodiment, the potential cause is cooling system failure, the cause feature of which is that the action object is the stator winding cooling system, the action mode is the cooling system flow reduction, the action timing is t1, and the expected impact effect is the temperature rise of the stator winding.

[0119] Step S14062: For each potential cause, locating the corresponding abnormal association node, and extracting the state change data of the abnormal association node before and after the abnormality, including the running parameter change, the scene feature change and the node interaction behavior change.

[0120] In this embodiment, the abnormality associated node corresponding to the potential cause is the stator winding temperature record node. The state change data of the node before and after the abnormality occurs is extracted. The operating parameter change is that the stator winding temperature rises from the normal range to the abnormal range. The scene feature change is that the stator winding temperature is unevenly distributed. The node interaction behavior change is that the stator winding temperature data transmission frequency increases.

[0121] Step S14063: Analyze the consistency of the device type and the scene attribute of the action object and the abnormality associated node in the cause feature, and compare the expected impact effect in the cause feature with the actual state change data of the abnormality associated node to identify the matching points and the difference points of the two.

[0122] In this embodiment, when analyzing the consistency of the device type and the scene attribute of the action object and the abnormality associated node, it is checked whether the action object is the device type of the abnormality associated node and whether the scene attribute matches. For example, the action object is the stator winding cooling system, and the device type of the abnormality associated node is the stator winding. The device types of the two are consistent, and the scene attributes match. When comparing the expected impact effect with the actual state change data, it is checked whether the expected impact effect is consistent with the actual state change data. For example, the expected impact effect is that the stator winding temperature rises, and the actual state change data is also that the stator winding temperature rises. The matching point of the two is the temperature rise, and the difference point is that the temperature rise amplitudes are different.

[0123] Step S14064: Based on the inter-node propagation logic in the abnormality propagation track, it is analyzed whether the potential cause can trigger the state change of the subsequent node and whether it matches the inter-node coupling association type. The multi-dimensional operating information corresponding to the potential cause action opportunity is extracted, and it is analyzed whether the operating environment and the device state under the potential cause action opportunity support the cause to play a role.

[0124] In this embodiment, when analyzing whether the potential cause can trigger the state change of the subsequent node, according to the inter-node propagation logic, it is checked whether the potential cause can cause the state change of the subsequent node. For example, the potential cause is a cooling system failure, which causes the temperature of the stator winding to rise, the temperature of the stator winding causes the guide vane opening to decrease, the guide vane opening causes the water flow velocity to decrease, and the water flow velocity causes the unit speed to decrease. The potential cause can trigger the state change of the subsequent node. When analyzing whether it matches the inter-node coupling association type, it is checked whether the coupling association type corresponding to the potential cause is consistent with the coupling association type between the nodes. For example, the coupling association type corresponding to the potential cause is an action-response type association, and the coupling association type between the nodes is also an action-response type association. They match. When extracting the multi-dimensional operation information corresponding to the action opportunity of the potential cause, the operation environment data at t1 is obtained, such as environmental temperature, humidity, etc., and the device state data is obtained, such as cooling system flow, pressure, etc. When analyzing whether the operation environment and the device state support the cause to play a role, it is checked whether the environmental temperature is too high, whether the cooling system flow is reduced, etc.

[0125] Step S14065: Construct a cause association evaluation framework, comprehensively evaluate from each dimension of action object consistency, influence effect fitness, propagation logic matching and action opportunity adaptability, extract dimension features for each evaluation dimension, generate an evaluation feature vector, input the evaluation feature vector into the evaluation module of the hydropower station operation coupling large model, comprehensively analyze the evaluation feature vector, generate a quantified association result, compare the quantified association result with a preset association standard, determine whether the potential cause passes the association verification, and record the verification process, evaluation feature vector and quantified result of each potential cause.

[0126] In this embodiment, the cause association evaluation framework includes the action object consistency dimension, the influence effect fitness dimension, the propagation logic matching dimension and the action opportunity adaptability dimension. The dimension features of the action object consistency dimension are the matching degree of the action object and the device type and scene attribute of the abnormal association node; the dimension features of the influence effect fitness dimension are the matching degree of the expected influence effect and the actual state change data; the dimension features of the propagation logic matching dimension are the matching degree of the potential cause and the inter-node propagation logic; and the dimension features of the action opportunity adaptability dimension are the matching degree of the action opportunity of the potential cause and the operation environment and device state. When generating the evaluation feature vector, the dimension features of each dimension are converted into numerical values and combined into a vector. When the evaluation module comprehensively analyzes the evaluation feature vector, a weighted summation method is used to weight process the numerical values of each dimension to generate a quantified association result. When the quantified association result is compared with the preset association standard, if the quantified result is greater than the preset standard, it is considered that the potential cause passes the association verification.

[0127] Step S14066: For the potential causes that do not pass the verification, analyze the key factors of verification failure to form verification feedback information. Based on the verification feedback information, recheck the abnormal association node state change data and propagation logic corresponding to the potential causes that do not pass the verification, confirm the rechecking result, finally determine the effective causes that pass the association verification, and output the effective causes that pass the association verification and the verification report.

[0128] In this embodiment, the potential cause that does not pass the verification is that the environmental temperature is too high, and the key factor of verification failure is that the matching degree of the action object consistency dimension is low, because the action object is the environmental temperature, and the device type of the abnormal association node is the stator winding, and the device types of the two are inconsistent. When forming the verification feedback information, the key factor that the matching degree of the action object consistency dimension is low is recorded. When rechecking, it is checked whether the abnormal association node state change data and the propagation logic are error, and after confirming the rechecking result, it is determined that the effective cause that passes the association verification is the cooling system failure. When outputting the verification report, the name, cause characteristics, verification process, evaluation feature vector and quantization result of the effective cause are recorded.

[0129] Step S1407: For each potential cause, perform cause association verification, analyze the fit of the potential cause and the abnormal node state change and the inter-node propagation logic, and take the potential cause that meets the requirement as the effective cause, and sort it according to its action order in the abnormal propagation track.

[0130] Step S1408: Build a cause association chain, take the effective cause of the initial deviation node as the starting point of the cause chain, take the effective cause of the intermediate propagation node as the intermediate link of the cause chain, and take the effective cause of the terminal impact node as the extension link of the cause chain.

[0131] In this embodiment, the effective cause of the initial deviation node is the cooling system failure, which is taken as the starting point of the cause chain; the effective cause of the intermediate propagation node is the guide vane opening control module failure, which is taken as the intermediate link of the cause chain; and the effective cause of the terminal impact node is the unit speed regulation system failure, which is taken as the extension link of the cause chain. The built cause association chain is cooling system failure→guide vane opening control module failure→unit speed regulation system failure.

[0132] Step S1409: Analyze the interaction relationship between the causes in each link of the cause chain, supplement the cause interaction logic, form the cause chain framework, input the cause chain framework into the cause chain optimization module of the hydropower station operation coupled large model, logically perfect the cause chain framework, and generate an abnormal cause chain.

[0133] For example, step S14091: input the cause chain framework into the cause chain optimization module of the hydropower station operation coupling large model, extract the cause characteristics, action objects and influence ranges of each cause node in the cause chain framework, analyze the logical relationship between adjacent cause nodes one by one, and identify the conflict links of cause action object conflict, influence range conflict and action mode conflict.

[0134] In this embodiment, the cause chain framework is cooling system failure → guide vane opening control module failure → unit speed regulation system failure. When extracting the cause characteristics, action objects and influence ranges of each cause node, the cause characteristics of the cooling system failure are cooling system flow reduction, the action object is the stator winding cooling system, and the influence range is the stator winding temperature; the cause characteristics of the guide vane opening control module failure are guide vane opening control signal abnormality, the action object is the guide vane opening control module, and the influence range is the guide vane opening; and the cause characteristics of the unit speed regulation system failure are unit speed regulation signal abnormality, the action object is the unit speed regulation system, and the influence range is the unit speed. When analyzing the logical relationship between adjacent cause nodes, it is checked whether the cooling system failure leads to the guide vane opening control module failure, and whether the guide vane opening control module failure leads to the unit speed regulation system failure. When identifying the conflict links, it is checked whether the cause action objects conflict, such as the action object of the cooling system failure being the stator winding cooling system and the action object of the guide vane opening control module failure being the guide vane opening control module, which do not conflict; whether the influence ranges conflict, such as the influence range of the cooling system failure being the stator winding temperature and the influence range of the guide vane opening control module failure being the guide vane opening, which do not conflict; and whether the action modes conflict, such as the action mode of the cooling system failure being flow reduction and the action mode of the guide vane opening control module failure being control signal abnormality, which do not conflict.

[0135] Step S14092: for each conflict link, retrieve the corresponding abnormal associated node operation information and propagation trajectory segment, analyze the root cause of the conflict, and based on the root cause analysis result, adjust the cause description or action range of the conflict cause node to eliminate the cause conflict.

[0136] In this embodiment, if there is a conflict link, such as the influence range of the cooling system failure conflicting with the influence range of the guide vane opening control module failure, the corresponding abnormal associated node operation information and propagation trajectory segment are retrieved, the root cause of the conflict is analyzed, such as the influence range of the cooling system failure being incorrectly expanded to the guide vane opening. When adjusting the cause description or action range of the conflict cause node, the influence range of the cooling system failure is corrected to the stator winding temperature to eliminate the cause conflict.

[0137] Step S14093: After the conflict processing is completed, the cause chain framework is traversed to identify a logical breakpoint of the cause association logic that is not coherent, for each logical breakpoint, a potential association clue of cause nodes before and after the logical breakpoint is found, based on the potential association clue, an intermediate cause node or cause interaction logic that can connect the breakpoint is supplemented, the potential association clue includes a common action object, a similar action mode and an associated abnormal node.

[0138] In this embodiment, when traversing the cause chain framework, if it is found that the cause association logic between the cooling system failure and the guide vane opening control module failure is not coherent, there is a logical breakpoint. When finding the potential association clue, it is found that the common action object of the two is the hydroelectric generating set, the similar action mode is to affect the equipment operation through the control signal, and the associated abnormal node is the stator winding temperature recording node. Based on these potential association clues, the intermediate cause node is supplemented to be that the stator winding temperature is too high to cause the guide vane opening control module to receive an error signal, or the cause interaction logic is supplemented to be that the cooling system failure causes the stator winding temperature to be too high, and the stator winding temperature that is too high causes the control signal of the guide vane opening control module to be abnormal.

[0139] Step S14094: The logical structure of a similar cause chain in the abnormal cause knowledge base is extracted, and the breakpoint connection of the current cause chain is perfected with reference to the association mode thereof.

[0140] In this embodiment, the logical structure of a similar cause chain in the abnormal cause knowledge base is extracted, such as the logical structure of cooling system failure→stator winding temperature that is too high→guide vane opening control module failure→guide vane opening abnormality→hydroelectric generating set speed abnormality. With reference to the logical structure, the breakpoint connection of the current cause chain is perfected, and the intermediate cause node of the stator winding temperature that is too high is supplemented, so that the cause chain framework becomes cooling system failure→stator winding temperature that is too high→guide vane opening control module failure→hydroelectric generating set speed regulation system failure.

[0141] Step S14010: The abnormal cause chain is associated with the abnormal propagation trajectory for the relevance verification, and the abnormal cause chain that passes the verification is output.

[0142] In this embodiment, the relevance verification adopts the trajectory matching method, the cause nodes in the abnormal cause chain are corresponded with the nodes in the abnormal propagation trajectory, and it is checked whether the action object, the action mode and the action timing of the cause nodes are consistent with the node state change and the interaction logic in the abnormal propagation trajectory. When the difference of the corresponding content is less than a preset threshold, it is considered that the verification passes. For example, the node corresponding to the cooling system failure in the abnormal cause chain is the stator winding temperature recording node, the action object, the action mode and the action timing thereof are consistent with the node state change and the interaction logic in the abnormal propagation trajectory, and the verification passes. When the abnormal cause chain that passes the verification is output, the cause nodes, the cause association logic, the action path and the influence result in the cause chain are combined together to form the structured cause chain data.

[0143] Step S150: generating a dynamically adapted hydropower station abnormal monitoring response scheme according to the abnormal propagation trajectory and the abnormal cause chain, and pushing the hydropower station abnormal monitoring response scheme to a hydropower station monitoring execution system.

[0144] Step S1501: analyzing a terminal influence node and an influence range of the abnormal propagation trajectory, and locating a hydropower station operation system, a device combination and an operation link affected by the abnormality.

[0145] In this embodiment, the terminal influence node is a unit speed record node, and the influence range is abnormal fluctuation of unit speed. When the abnormality is located to affect the hydropower station operation system, it is determined to be a hydro-generator set operation system; when the abnormality is located to affect the device combination, it is determined to be a hydro-generator set, a guide vane opening control module, a cooling system and the like; and when the abnormality is located to affect the operation link, it is determined to be a unit start-up link, a unit operation link, a unit shutdown link and the like.

[0146] Step S1502: analyzing a core cause node, cause correlation logic and an action path in the abnormal cause chain, and determining a key trigger factor and a propagation boost factor of the abnormality.

[0147] In this embodiment, the core cause node is a cooling system fault, the cause correlation logic is that the cooling system fault causes the stator winding temperature to be too high, the stator winding temperature being too high causes the guide vane opening control module to be faulty, and the guide vane opening control module being faulty causes the unit speed to be abnormal, and the action path is cooling system→stator winding→guide vane opening control module→unit speed regulation system. The key trigger factor is the cooling system fault, and the propagation boost factor is that the environmental temperature being too high aggravates the influence of the cooling system fault.

[0148] Step S1503: calling a preset hydropower station abnormal response strategy library, the hydropower station abnormal response strategy library including response measures, execution processes and adaptation conditions corresponding to different cause types, influence ranges and propagation speeds.

[0149] In this embodiment, the hydropower station abnormal response strategy library includes response measures corresponding to the cooling system fault, such as starting a backup cooling system, reducing unit output, shutdown maintenance and the like; includes response measures corresponding to the influence range of abnormal fluctuation of unit speed, such as adjusting the guide vane opening, adjusting the paddle opening, switching the power transmission line and the like; and includes response measures corresponding to the fast propagation speed, such as emergency shutdown, cutting off the power transmission line and the like.

[0150] Step S1504: according to the type and the influence range of the core cause node, screening a preliminarily adapted response measure set from the hydropower station abnormal response strategy library.

[0151] In this embodiment, the type of the core cause node is cooling system failure, the influence range is abnormal fluctuation of unit rotation speed, and the set of preliminary adapted response measures selected from the abnormal response strategy library of the hydropower station is starting the standby cooling system, reducing unit output, and adjusting guide vane opening.

[0152] Step S1505: In combination with the propagation speed of the abnormal propagation trajectory and the node interaction intensity, the preliminary adapted response measures are prioritized, the adaptation conditions of each response measure are analyzed in combination with the fitting situation of the current hydropower station operation state and equipment load condition, the specific execution parameters of the response measures are adjusted, the adjusted response measures are bound with the key nodes in the abnormal propagation trajectory, and the action nodes, execution time and expected effect corresponding to each response measure are determined.

[0153] In this embodiment, the propagation speed of the abnormal propagation trajectory is fast, and the node interaction intensity is large. When prioritizing the preliminary adapted response measures, starting the standby cooling system is ranked first, reducing unit output is ranked second, and adjusting the guide vane opening is ranked third. When analyzing the adaptation conditions of each response measure in combination with the fitting situation of the current hydropower station operation state and equipment load condition, it is checked whether the adaptation conditions of starting the standby cooling system are met, such as whether the standby cooling system is in a normal state, whether the current hydropower station operation state allows starting the standby cooling system, and whether the equipment load condition allows reducing unit output. When adjusting the specific execution parameters of the response measures, the flow of starting the standby cooling system is adjusted to 1.2 times the normal flow, the amplitude of reducing unit output is adjusted to 20%, and the amplitude of adjusting the guide vane opening is adjusted to 10%. The adjusted response measures are bound with the key nodes in the abnormal propagation trajectory, such as the action node of starting the standby cooling system is the cooling system, the execution time is t2 moment, and the expected effect is to reduce the temperature of the stator winding; the action node of reducing unit output is the hydro-generator unit, the execution time is t3 moment, and the expected effect is to stabilize the unit rotation speed; and the action node of adjusting the guide vane opening is the guide vane opening control module, the execution time is t4 moment, and the expected effect is to restore the guide vane opening to normal.

[0154] Step S1506: In view of the indirect causes and cause interaction relationship in the abnormal cause chain, corresponding auxiliary response measures are supplemented to form a response measure system.

[0155] In this embodiment, the indirect cause in the abnormal cause chain is that the environmental temperature is too high, and the cause interaction relationship is that the environmental temperature is too high to aggravate the influence of the cooling system failure. The supplementary auxiliary response measures are to start the plant ventilation system, reduce the environmental temperature, or adjust the reservoir water level to reduce the water flow temperature. When forming the response measure system, the preliminary adapted response measures and the auxiliary response measures are combined together to form a complete response measure system.

[0156] Step S1507: constructing an execution flow of the response scheme, arranging the response measures in priority order, determining the connection logic, execution order and coordination mode among the response measures, assigning a corresponding execution subject identifier to each response measure, determining the execution department, operation post and monitoring responsibility subject.

[0157] In this embodiment, the execution flow arranges the response measures in priority order as starting the standby cooling system → reducing the unit output → adjusting the guide vane opening → starting the plant ventilation system. The connection logic is that after starting the standby cooling system, if the stator winding temperature is not reduced, the unit output is reduced; after reducing the unit output, if the unit speed is still unstable, the guide vane opening is adjusted; after adjusting the guide vane opening, if the environmental temperature is still too high, the plant ventilation system is started. The execution order is that the standby cooling system is started at time t2, the unit output is reduced at time t3, the guide vane opening is adjusted at time t4, and the plant ventilation system is started at time t5. The coordination mode is that the execution subject of starting the standby cooling system and the execution subject of reducing the unit output keep communication and timely feedback equipment state information. A corresponding execution subject identifier is assigned to each response measure, such as the execution subject identifier of starting the standby cooling system is the operation and maintenance department A, the execution subject identifier of reducing the unit output is the operation department B, the execution subject identifier of adjusting the guide vane opening is the control department C, and the execution subject identifier of starting the plant ventilation system is the logistics department D. The execution department is determined as the operation and maintenance department, the operation department, the control department and the logistics department, the operation post is determined as the operation and maintenance engineer, the operation operator, the control engineer and the logistics personnel, and the monitoring responsibility subject is determined as the monitoring center.

[0158] Step S1508: obtaining a response effect evaluation standard, the response effect evaluation standard including evaluation dimensions of abnormality mitigation degree, node state recovery situation and propagation trajectory blocking effect.

[0159] In this embodiment, the evaluation dimensions of the abnormality mitigation degree are the amplitude of the stator winding temperature reduction, the degree of unit speed stabilization, etc.; the evaluation dimensions of the node state recovery situation are the time for the stator winding temperature to recover to the normal range, the time for the guide vane opening to recover to the normal range, etc.; and the evaluation dimensions of the propagation trajectory blocking effect are whether the abnormal propagation trajectory is blocked, whether subsequent nodes no longer appear deviation, etc.

[0160] Step S1509: structurally integrate the response measure system, the execution process, the execution subject identifier, and the response effect evaluation standard, generate an initial response scheme, input the initial response scheme into the scheme optimization module of the hydropower station operation coupled large model, dynamically optimize the initial response scheme, simulate the running state change after the execution of the response scheme, adjust the execution parameters, execution time, and coordination logic of the response measures based on the simulation results, improve the adaptability and effectiveness of the scheme, and generate a final hydropower station abnormal monitoring response scheme. The hydropower station abnormal monitoring response scheme includes core response measures, auxiliary response measures, an execution process, an execution subject, evaluation standards, and an emergency adjustment plan.

[0161] In this embodiment, when structurally integrating, the response measures in the response measure system are arranged in priority order, the connection logic, execution order, and coordination mode in the execution process are matched with the response measures, the execution subject identifier is matched with the response measures, and the response effect evaluation standard is matched with the expected effect of the response measures, to form an initial response scheme. When the scheme optimization module dynamically optimizes the initial response scheme, the running state change after the execution of the response scheme is simulated, such as the change in the temperature of the stator winding after starting the standby cooling system, the change in the speed of the unit after reducing the unit output, etc. Based on the simulation results, the execution parameters of the response measures are adjusted, such as adjusting the flow of starting the standby cooling system to 1.3 times the normal flow; the execution time is adjusted, such as advancing the execution time of reducing the unit output to t2.5; the coordination logic is adjusted, such as strengthening the real-time data transmission between the execution subjects of starting the standby cooling system and reducing the unit output. After improving the adaptability and effectiveness of the scheme, a final hydropower station abnormal monitoring response scheme is generated, including core response measures, auxiliary response measures, an execution process, an execution subject, evaluation standards, and an emergency adjustment plan. The emergency adjustment plan is a backup response measure taken when the execution effect of the response measures is not good, such as emergency shutdown, cutting off the power transmission line, etc.

[0162] Step S15010: push the hydropower station abnormal monitoring response scheme to the hydropower station monitoring execution system.

[0163] In this embodiment, the push process adopts a message queue mode, converts the hydropower station abnormal monitoring response scheme into a message format, and sends it to the message queue of the hydropower station monitoring execution system. The hydropower station monitoring execution system obtains the response scheme from the message queue, parses the content of the response scheme, and executes the response measures according to the execution process.

[0164] In one exemplary embodiment, a hydropower station abnormal monitoring system based on large model data analysis is provided, which can be a terminal, a server, etc., and its internal structure diagram can be as follows Figure 2The abnormal monitoring system of the hydropower station based on large model data analysis shown in the figure includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through the system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor is used to provide computing and control capability. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface is used to exchange information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals, and wireless communication can be realized through WIFI, mobile cellular network, near field communication or other technologies. The computer program is executed by the processor to realize an abnormal monitoring method of a hydropower station based on large model data analysis. The display unit is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device can be a touch layer overlaid on the display screen, or a button, trackball or touchpad arranged on the shell of the abnormal monitoring system of the hydropower station based on large model data analysis, or an external keyboard, touchpad or mouse, etc.

[0165] It should be noted that, in order to simplify the description of the present disclosure and help understand one or more embodiments of the present disclosure, in the foregoing description of embodiments of the present disclosure, various features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A hydropower station anomaly monitoring method based on large model data analysis, characterized in that, The method comprises: Collecting multi-dimensional running information generated by the whole scene operation of the hydropower station, wherein the multi-dimensional running information comprises equipment working condition information, water flow dynamic information, power transmission information and environment interaction information; Performing cross-scene coupling correlation modeling on the multi-dimensional running information by using a pre-trained hydropower station operation coupling large model to generate an operation state coupling correlation model; Based on the operation state coupling correlation model, combining with normal operation benchmark data of the hydropower station, reconstructing an abnormal propagation trajectory in the operation process of the hydropower station; Using the hydropower station operation coupling large model to trace the cause chain of the coupling correlation nodes involved in the abnormal propagation trajectory to generate an abnormal cause chain; According to the abnormal propagation trajectory and the abnormal cause chain, generating a dynamically adapted hydropower station abnormal monitoring response scheme, and pushing the hydropower station abnormal monitoring response scheme to a hydropower station monitoring execution system; The method comprises: Collecting multi-dimensional historical running information accumulated during a long-term normal operation stage of the hydropower station; Splitting the multi-dimensional historical running information according to scene types to form a historical equipment working condition information set, a historical water flow dynamic information set, a historical power transmission information set and a historical environment interaction information set; Extracting historical scene features of running records in each historical information set, wherein the historical scene features of the historical equipment working condition information set comprise historical equipment running parameter change features and historical equipment cooperation correlation features, the historical scene features of the historical water flow dynamic information set comprise historical water flow pattern change features and historical water flow and equipment interaction features, the historical scene features of the historical power transmission information set comprise historical transmission efficiency change features and historical transmission link correlation features, and the historical scene features of the historical environment interaction information set comprise historical environmental factor change features and historical environment and equipment interaction features; Identifying historical coupling correlation dimensions between different historical information sets, wherein the historical coupling correlation dimension between the historical equipment working condition information set and the historical water flow dynamic information set is a historical action response dimension, the historical coupling correlation dimension between the historical equipment working condition information set and the historical power transmission information set is a historical load adaptation dimension, the historical coupling correlation dimension between the historical water flow dynamic information set and the historical environment interaction information set is a historical influence feedback dimension, and the historical coupling correlation dimension between the historical power transmission information set and the historical environment interaction information set is a historical adaptation adjustment dimension; For each historical coupling correlation dimension, a corresponding historical coupling correlation factor is generated, wherein the historical coupling correlation factor is determined based on the interaction relationship between the historical scene features in different historical information sets; Input all historical coupling correlation factors into an association modeling module of the hydropower station operation coupling large model to construct a historical association logic network between the multi-dimensional historical running information; In the historical association logic network, each historical running record is taken as a historical network node, and each historical coupling correlation factor is taken as a connection link between historical nodes to form a historical initial coupling correlation framework; The dynamic optimization module of the hydropower station operation coupling large model iteratively optimizes the historical initial coupling correlation framework, adjusts the correlation strength and correlation logic between historical nodes, and eliminates historical correlation conflicts and historical redundant correlations; Integrate the optimized historical correlation logic network and the historical scene characteristics of each historical information set to generate multiple normal operation state coupling correlation models, extract the common correlation logic, stable coupling factors and node interaction rules in all normal operation state coupling correlation models, and integrate them to form the hydropower station normal operation benchmark data; Comprehensively compare the current generated operation state coupling correlation model with the hydropower station normal operation benchmark data to identify the deviation correlation nodes and deviation coupling factors in the operation state coupling correlation model that deviate from the normal operation benchmark data; Trace the original multi-dimensional operation information corresponding to each deviation correlation node to extract the operation state characteristics and correlation interaction characteristics of the deviation correlation node; Analyze the coupling correlation relationship between the deviation correlation nodes, and deduce the connection path and interaction sequence between the deviation correlation nodes based on the correlation logic network in the operation state coupling correlation model; Identify the coupling correlation dimension corresponding to the deviation coupling factor, and determine the propagation direction and influence range of the deviation in different coupling correlation dimensions; Based on the connection path, interaction sequence and deviation propagation direction of the deviation correlation node, an initial abnormal propagation path framework is constructed; In the initial abnormal propagation path framework, after supplementing the deviation degree, interaction time sequence and coupling factor change of each node, the initial abnormal propagation path framework is input into the trajectory reconstruction module of the hydropower station operation coupling large model for logical verification and structural optimization, and the node interaction logic in the path is adjusted to be consistent with the coupling correlation model; According to the optimized path framework, the abnormal propagation trajectory is reconstructed, which includes the initial deviation node, the intermediate propagation node, the terminal impact node and the propagation logic chain between nodes; Each propagation link in the abnormal propagation trajectory is labeled, the coupling correlation dimension, deviation trend and node interaction mode of each link are determined, and the labeled abnormal propagation trajectory is output; The use of the hydropower station operation coupling large model to trace the cause chain of the coupling correlation nodes involved in the abnormal propagation trajectory generates an abnormal cause chain, including: Extract all coupling correlation nodes in the abnormal propagation trajectory to form an abnormal correlation node set, including the initial deviation node, the intermediate propagation node and the terminal impact node; Collect the multi-dimensional operation information original records and scene characteristics corresponding to each abnormal correlation node to form a node correlation information set; Input the abnormal correlation node set and the node correlation information set into the cause tracing module of the hydropower station operation coupling large model, call the preset hydropower station abnormal cause knowledge base, and the hydropower station abnormal cause knowledge base contains the abnormal correlation nodes, propagation trajectory fragments and complete cause records corresponding to historical abnormal events. The cause record includes direct cause, indirect cause and cause interaction relationship; The semantic matching module of the hydropower station operation coupling large model is used for deep matching of the scene features of the current abnormal associated node and the scene features of the historical abnormal associated node in the knowledge base, screening of the historical abnormal events with the scene feature matching degree meeting the requirements, extraction of the corresponding cause records and propagation track segments, and comparison and analysis of the historical cause records with the node interaction logic and coupling association type of the current abnormal propagation track. The reusable cause association clues are identified based on the cause association clues, and the potential causes corresponding to each abnormal associated node are preliminarily inferred in combination with the original records of the operation information of the current abnormal associated node. The cause association of each potential cause is verified, the fitting condition of the potential cause and the abnormal node state change and the inter-node propagation logic is analyzed, the potential cause meeting the fitting condition requirement is taken as an effective cause, and the effective cause is sorted according to the action order in the abnormal propagation track. A cause association chain is constructed, the effective cause of the initial bias node is taken as a starting point of the cause chain, the effective cause of the intermediate propagation node is taken as an intermediate link of the cause chain, and the effective cause of the terminal influence node is taken as an extension link of the cause chain. The interaction relationship between the causes in each link of the cause chain is analyzed, the cause interaction logic is supplemented, a cause chain framework is formed, the cause chain framework is input into a cause chain optimization module of the hydropower station operation coupling large model, the cause chain framework is logically perfected, and in the optimized cause chain, the abnormal associated node, the action mode and the influence range corresponding to each cause are labeled to generate a structured abnormal cause chain, the abnormal cause chain includes a cause node, a cause association logic, an action path and an influence result. The abnormal cause chain is associated with the abnormal propagation track for correlation verification, and the abnormal cause chain passing the verification is output.

2. The hydropower station abnormality monitoring method based on large model data analysis according to claim 1, characterized in that, The pre-trained hydropower station operation coupling large model is used for cross-scene coupling association modeling of the multi-dimensional operation information to generate an operation state coupling association model, including: The multi-dimensional operation information is split according to scene types to form a device working condition information set, a water flow dynamic information set, a power transmission information set and an environment interaction information set, each information set containing a plurality of specific operation records under the same scene; Scene features of the operation records in each information set are extracted, the scene features of the device working condition information set include device operation parameter change features and device cooperation association features, the scene features of the water flow dynamic information set include water flow pattern change features and water flow and device action features, the scene features of the power transmission information set include transmission efficiency change features and transmission link association features, and the scene features of the environment interaction information set include environment factor change features and environment and device action features; Coupling association dimensions between different information sets are identified, the coupling association dimensions of the device working condition information set and the water flow dynamic information set are action response dimensions, the coupling association dimensions of the device working condition information set and the power transmission information set are load adaptation dimensions, the coupling association dimensions of the water flow dynamic information set and the environment interaction information set are influence feedback dimensions, and the coupling association dimensions of the power transmission information set and the environment interaction information set are adaptation adjustment dimensions; For each coupling correlation dimension, scene features in the corresponding two information sets are extracted, interaction relationships between the scene features are analyzed, feature indicators capable of representing interaction strength or adaptation degree are extracted, and a feature indicator set corresponding to each dimension is formed; Each feature indicator set is subjected to redundant information elimination processing, core feature indicators capable of reflecting core interaction relationships are retained, and all core feature indicators are input into a factor generation module of the hydropower station operation coupling large model to generate coupling correlation factors corresponding to each coupling correlation dimension, which are directly related to the core feature indicators of the corresponding dimension; All coupling correlation factors are input into an association modeling module of the hydropower station operation coupling large model, which constructs an association logic network between multi-dimensional operation information based on the deep semantic understanding capability of the large model, in which each operation record is taken as a network node and each coupling correlation factor is taken as a connection link between nodes to form an initial coupling association framework; Through a dynamic optimization module of the hydropower station operation coupling large model, the initial coupling association framework is iteratively optimized, the association strength and association logic between nodes are adjusted, the optimized association logic network and the scene features of each information set are integrated, and an operation state coupling association model capable of reflecting the cross-scene interaction relationships of multi-dimensional operation information is generated; The operation state coupling association model is subjected to scene adaptation adjustment, and an adjusted operation state coupling association model is output.

3. The hydropower station abnormality monitoring method based on large model data analysis according to claim 1, characterized in that, The coupling association relationships between the deviation association nodes are analyzed, and the connection path and interaction sequence between the deviation association nodes are derived based on the association logic network in the operation state coupling association model, including: The node attributes and association identifiers of all deviation association nodes are extracted from the association logic network of the operation state coupling association model; Based on the node attributes and association identifiers, the position and adjacent association nodes of each deviation association node in the association logic network are determined, including directly associated normal nodes and deviation nodes; The coupling association types between each deviation association node and adjacent association nodes are analyzed, including action response type association, load adaptation type association, influence feedback type association and adaptation adjustment type association; For each deviation association node, the interaction history between it and adjacent association nodes is traced, and data transmission records, state change records and coupling factor change records in the interaction process are extracted; Based on the interaction history records, the interaction cause-effect relationship between the deviation association nodes and adjacent association nodes is derived, and the trigger condition and transmission logic of the deviation propagation are determined; Starting from the initial deviation node, the deviation nodes directly associated with the initial deviation node are sequentially derived according to the cause-effect relationship and coupling association type to form a first-level propagation path; Taking the deviation nodes in the first-level propagation path as the starting point, the subsequent deviation nodes associated with them are continuously derived to form a second-level propagation path, and the process is repeated in this way until all associated deviation nodes are covered, the node sequence, coupling association type and interaction time node in each propagation path are recorded, and a preliminary path sequence is formed; The initial path sequence is deduplicated. The interaction order of each node in the deduplicated path sequence is analyzed to see if it matches the preset interaction rules in the associated logic network. The order of nodes that do not match is adjusted, and all path sequences that match the rules are integrated to form a complete connection path and interaction order between the deviated associated nodes. The complete connection path and interaction sequence are mapped to the associated logical network, and the specific location and relationship of the path in the network are marked. Based on the marked mapping results, the accuracy of the connection path and interaction sequence is checked again, the path deviation found in the mapping process is corrected, and the deduced connection path and interaction sequence are output.

4. The hydropower station abnormality monitoring method based on large model data analysis according to claim 1, characterized in that, The process of verifying the causal correlation of each potential cause, and analyzing the fit between the potential cause and the changes in the state of abnormal nodes and the propagation logic between nodes, includes: Extract the causal features corresponding to each potential cause. The causal features include the object of the cause, the mode of action, the timing of action, and the expected impact. For each potential cause, locate its corresponding abnormal associated node, and extract the state change data of the abnormal associated node before and after the anomaly occurs. The state change data includes changes in operating parameters, changes in scene features, and changes in node interaction behavior. Analyze the consistency between the target device and the device type and scene attributes of the abnormal associated nodes in the causal characteristics, and compare the expected impact effect in the causal characteristics with the actual state change data of the abnormal associated nodes to identify the points of convergence and differences between the two. Based on the inter-node propagation logic in the abnormal propagation trajectory, we analyze whether the potential cause can trigger the state changes of subsequent nodes and whether it matches the type of coupling and association between nodes. We extract multi-dimensional operational information corresponding to the timing of the potential cause's action and analyze whether the operating environment and equipment status under the timing of the potential cause's action support the cause's action. A causal correlation assessment framework is constructed, which comprehensively evaluates the correlation from various dimensions, including consistency of the target, fit of the impact effect, matching of the propagation logic, and suitability of the timing of action. Dimensional features are extracted for each assessment dimension to generate assessment feature vectors. The assessment feature vectors are input into the assessment module of the large-scale coupled model of the hydropower station operation, and the assessment feature vectors are comprehensively analyzed to generate correlation quantification results. The correlation quantification results are compared with preset correlation standards to determine whether the potential causes have passed the correlation verification. The verification process, assessment feature vectors, and quantification results of each potential cause are recorded. For potential causes of failure to pass verification, analyze the key factors of verification failure, generate verification feedback information, and based on the verification feedback information, re-examine the abnormal related node status change data and propagation logic corresponding to the potential causes of failure to pass verification. After confirming the verification results, finally determine the valid causes of passing the correlation verification, and output the valid causes of passing the correlation verification and the verification report.

5. The hydropower station anomaly monitoring method based on large model data analysis according to claim 1, characterized in that, The process of generating a dynamically adapted hydropower station anomaly monitoring and response scheme based on the anomaly propagation trajectory and anomaly cause chain includes: Analyze the terminal impact nodes and impact range of the abnormal propagation trajectory, and locate the hydropower station operation system, equipment combination and operation links affected by the anomaly; Analyzing the core cause nodes, cause correlation logic and action paths in the abnormal cause chain to determine the key trigger factors and propagation boost factors of the abnormality; Calling a preset hydropower station abnormal response strategy library, which includes response measures, execution processes and adaptation conditions corresponding to different cause types, influence ranges and propagation speeds; According to the type and influence range of the core cause node, a preliminary set of response measures is selected from the hydropower station abnormal response strategy library; The response measures are prioritized by combining the propagation speed and node interaction intensity of the abnormal propagation trajectory, analyzing the adaptation conditions of each response measure to the current hydropower station operating state and equipment load situation, adjusting the specific execution parameters of the response measures, binding the adjusted response measures to the key nodes in the abnormal propagation trajectory, and determining the action nodes, execution timing and expected effects of each response measure; For indirect causes and cause interaction relationships in the abnormal cause chain, corresponding auxiliary response measures are supplemented to form a response measure system; An execution process of the response scheme is constructed, the response measures are arranged in priority order, the connection logic, execution order and coordination mode between the response measures are determined, and each response measure is assigned a corresponding execution subject identifier to determine the execution department, operating post and monitoring responsibility subject; Response effect evaluation criteria are obtained, which include evaluation dimensions such as abnormality mitigation degree, node state recovery situation and propagation trajectory blocking effect; The response measure system, execution process, execution subject identifier and response effect evaluation criteria are structurally integrated to generate an initial response scheme, which is input into the scheme optimization module of the hydropower station operation coupled large model for dynamic optimization of the initial response scheme, simulation of the operating state changes after execution of the response scheme, adjustment of the execution parameters, execution timing and coordination logic of the response measures based on the simulation results, improvement of the adaptability and effectiveness of the scheme, generation of the final hydropower station abnormal monitoring response scheme, and determination of the core response measures, auxiliary response measures, execution process, execution subject, evaluation criteria and emergency adjustment plan. The hydropower station abnormal monitoring response scheme is associated with the abnormal propagation trajectory and the abnormal cause chain, and after passing the association verification, the hydropower station abnormal monitoring response scheme is converted into a standardized execution instruction and pushed to the corresponding hydropower station monitoring execution system, and the hydropower station abnormal monitoring response scheme, abnormal propagation trajectory and abnormal cause chain are stored in the hydropower station abnormal treatment archive library to form an abnormal monitoring treatment record.

6. The hydropower station anomaly monitoring method based on large model data analysis according to claim 2, characterized in that, For each coupling correlation dimension, scene features are extracted from the corresponding two information sets, the interaction relationship between the scene features is analyzed, a feature index representing the interaction intensity or adaptation degree is extracted, and a feature index set corresponding to each dimension is formed, including: For the action response dimension, the characteristics of equipment operating parameter changes in the equipment condition information set and the characteristics of water flow morphology changes in the water flow dynamic information set are extracted. The interaction between the characteristics of equipment operating parameter changes and the characteristics of water flow morphology changes is analyzed. The influence of equipment operation on water flow morphology and the reaction of water flow morphology on equipment operation are identified. Based on the influence and reaction methods, feature indicators that can characterize the intensity of the interaction between the two are extracted to form a set of action response feature indicators. For the load adaptation dimension, we extract the equipment collaboration correlation features from the equipment operating condition information set and the transmission efficiency change features from the power transmission information set, analyze the adaptation relationship between the equipment collaboration correlation features and the transmission efficiency change features, identify the impact law of equipment collaboration mode on transmission efficiency and the adjustment requirements of transmission efficiency change on equipment collaboration, and extract feature indicators that can characterize the degree of adaptation between the two based on the impact law and the adjustment requirements, forming a load adaptation feature indicator set. For the feedback dimension, the characteristics of water flow and equipment interaction in the dynamic water flow information set and the characteristics of environment and equipment interaction in the environmental interaction information set are extracted. The feedback relationship between the characteristics of water flow and equipment interaction and the characteristics of environment and equipment interaction is analyzed. The influence of environmental factors on water flow state and the feedback effect of water flow state changes on environmental effects are identified. Based on the influence and feedback effect, feature indicators that can characterize the feedback intensity of both are extracted to form a set of influence feedback feature indicators. For the adaptation and adjustment dimension, the transmission link association features in the power transmission information set and the environmental factor change features in the environmental interaction information set are extracted. The adjustment relationship between the transmission link association features and the environmental factor change features is analyzed. The impact of environmental changes on the transmission link and the adaptation strategy of the transmission link to environmental changes are identified. Based on the impact and adaptation strategy, feature indicators that can characterize the degree of adaptation of the two are extracted to form a set of adaptation and adjustment feature indicators. The system integrates the action response characteristic index set, retaining core characteristic indicators; it filters the load adaptation characteristic index set, retaining indicators that reflect core adaptation relationships; it optimizes the impact feedback characteristic index set, strengthening the representation of key feedback relationships; and it integrates the adaptation adjustment characteristic index set, retaining indicators that reflect core adjustment adaptation relationships. Finally, it unifies the format of the processed action response characteristic index set, load adaptation characteristic index set, impact feedback characteristic index set, and adaptation adjustment characteristic index set, outputting a unified set of characteristic indicators for each dimension.

7. The hydropower station anomaly monitoring method based on large model data analysis according to claim 1, characterized in that, The process of inputting the initial anomaly propagation path framework into the trajectory reconstruction module of the coupled large-scale hydropower station operation model, performing logical verification and structural optimization on the initial anomaly propagation path framework, and adjusting the node interaction logic in the path to be consistent with the coupled correlation model includes: The initial abnormal propagation path framework is input into a trajectory reconstruction module of the hydropower station operation coupled large model, associated logic rules in an operation state coupled correlation model are extracted, including node interaction type rules, coupled correlation dimension rules and propagation order rules, whether each node interaction link in the initial path framework conforms to the associated logic rules is checked one by one, and abnormal interaction links that do not conform to the rules are identified; For each abnormal interaction link, node attributes, coupled correlation factors and interaction time information of the abnormal interaction link are extracted, reasons why the abnormal interaction link does not conform to the rules are analyzed, node interaction modes or interaction sequences are adjusted in combination with node correlation relationships in the associated logic network, application of the coupled correlation factors in the initial path framework is checked whether it matches the coupled correlation dimension, coupled correlation factors that do not match are adjusted, whether the propagation order of the nodes in the initial path framework conforms to actual change time sequences of multi-dimensional operation information is checked, and node sequences that are out of sequence are corrected, after the logical verification is completed, node distribution density of the adjusted path framework is analyzed, node links with density greater than a set density threshold are merged, the optimized path framework is mapped and compared with the operation state coupled correlation model, based on the mapping and comparison results, the path framework is finally adjusted, and an abnormal propagation trajectory is output. Comprise:

8. A hydropower station abnormal monitoring system based on large model data analysis, characterized in that, a processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the machine-executable instructions to perform the hydropower station abnormal monitoring method based on large model data analysis of any one of claims 1 to 7. ​

Images